Entertainment & Media

ABITS: An Agent Based Intelligent Tutoring System for Distance Learning

Description
ABITS: An Agent Based Intelligent Tutoring System for Distance Learning Nicola Capuano, Marco Marsella, Saverio Salerno CRMPA, Centro di Ricerca in Matematica Pura ed Applicata c/o DIIMA, University of
Published
of 11
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
ABITS: An Agent Based Intelligent Tutoring System for Distance Learning Nicola Capuano, Marco Marsella, Saverio Salerno CRMPA, Centro di Ricerca in Matematica Pura ed Applicata c/o DIIMA, University of Salerno, via Ponte Don Melillo, 84084, Fisciano (SA), Italy DIIMA, Dipartimento di Ingegneria dell Informazione e Matematica Applicata, University of Salerno, Via Ponte Don Melillo, 84084, Fisciano (SA), Italy Abstract. The purpose of this paper is to describe an Intelligent Tutoring Framework highly re-usable and suitable to several knowledge domains. In particular the system, named ABITS, has been realized in the context of the InTraSys ESPRIT project. It is able to support a Web-based Course Delivery Platform with a set of intelligent functions providing both student modeling and automatic curriculum generation. Such functions found their effectiveness on a set of rules for knowledge indexing based on Metadata and Conceptual Graphs following the IEEE Learning Object Metadata (LOM) standard. Moreover, in order to ensure the maximal flexibility, ABITS is organized as a Multi Agent System (MAS) composed by pools of three different kind of agents (evaluation, pedagogical and affective agents). Each agent is able to solve in autonomous way a specific task and they work together in order to improve the WBT learning effectiveness adapting the didactic materials to user skills and preferences. 1 Introduction An efficient Training System should allow users to take a lesson without time and place constrains. In order to fulfil these requirements, at present days, the best solution has to be naturally Web based. It requires to end-users zero cost installation and provides them the maximum time/place flexibility. Three kinds of Web-Based Tutoring (WBT) methodologies are available on the scene at this moment. Static WBT: teachers arrange learning material in order to cover one or more topics and convert them in interactive linked HTML pages (or different kinds of Web-deliverable objects). Material is then placed on-line in order to make it visible to everybody. Learners can exploit it only by following the path established by teachers. Personalized WBT: teachers, using a specific kind of software named Course Management System (i.e. Macromedia Attain) are able to perform manually a set of additional tasks. They can monitor student knowledge by testing them, assign recovery material if necessary, define different paths through learning objects for different kind of learning goals, etc. Adaptive WBT: includes all features of a Personalized WBT but the teacher is supported/simulated in his activity by using Artificial Intelligence techniques. In this paper we will present ABITS: an highly re-usable Intelligent Tutoring Framework able to extend a traditional Course Management System (CMS) with a set of intelligent functions allowing both student modeling and automatic curriculum generation. Adding ABITS as a module, any Personalized WBT will be able to become an Adaptive one. The requirement for the CMS is only one: it must be able to be extended with a scripting language supporting RMI invocation and able to access external data sources. Macromedia Attain, for example, fulfills these requirements. ABITS (Agent Based Intelligent Tutoring System) has been thought and developed in the context of the InTraSys ESPRIT project. InTraSys (Intelligent Training System in Technical Assistance) is a very complex training and learning system geared towards high-tech organizations whose objectives are to improve the training and learning effectiveness, reduce the training costs, increase industrial intellectual capital retention and decrease the employee training time [5]. The already quoted ABITS intelligent functions are summarized in the use case diagram of figure 1 and will be described in the following chapters. In particular, the Evaluate Curriculum function is dealt with curriculum sequencing and will be detailed in chapter 4; Evaluate Preferences and Evaluate Cognitive State concern, instead, user modeling and will be fully described in chapter 3. Evaluate Curriculum uses uses CMS Evaluate Preferences Evaluate All uses Evaluate Cognitive State Fig. 1. ABITS Use Case Diagram Such functions found their effectiveness on a set of rules for knowledge indexing based on Metadata and Conceptual Graphs. In the following chapter we will deepen this topic. 2 Knowledge Indexing Inside ABITS, all didactic material is organized in several Learning Objects and stored in a Course Material File System. A Learning Object is defined [1] as any entity which can be used, re-used or referenced during technology-supported learning. In our case, a Learning Object is a logical container that represents an atomic Webdeliverable resource such as a Lesson (an HTML page), a Simulation (a Java applet), a Virtual World (a VRML file), a Test (an HTML page with an evaluating form) and each kind of Web-deliverable object. Learning Objects must be indexed in order to let ABITS know what each one of them is about and how it can be used during the learning process. Some kind of information about Learning Objects is so required. We call this kind of information Metadata. 2.1 Metadata Metadata is information about an object, be it physical or digital and its main goal is to locate in efficient and effective way resources over a system or a computer network [2]. In the field of learning materials, several organizations such as IEEE, EDUCOM etc. have focused their attention on the creation of metadata standards specifying the syntax and the semantics of the so-called Learning Object Metadata. A Learning Object Metadata standard defines the minimal set of properties needed to allow these objects to be managed, located, and evaluated. It accommodates moreover the ability for locally extending the basic properties. Many advantages come from referring to a Learning Object Metadata standard: To take advantage of a complete syntax and semantic already created by experts of the Learning Technology. To enable the automatic importation of extern learning objects that adopt the same Metadata description standard. To enable the exportation/sale of learning objects created for ABITS to extern systems/clients that adopt the same Metadata description standard. We chose to adopt for ABITS the IEEE LTSC Learning Object Metadata (LOM) standard [1]. We seen, in fact, that other organization are slowly converging to this one: for the future it can be the best choice. LOM Metadata is structured in a hierarchical way: schemes consist of data elements. The latter are defined through: sub-schemes if they are a collection of data elements themselves; data types if their values are strings, decimals, etc; vocabularies if their values come from an enumerated list. Data elements moreover can be mandatory (must be present) or optional (may be present). IEEE metadata definition implies that descriptors of a learning resources are grouped in meaningful categories. Our schema proposes six categories: a subset of the eight defined by IEEE standard: General : groups all context independent features plus the semantic descriptors for the resource. Life Cycle: groups the features linked to the lifecycle of the resource. Meta Metadata: groups the features of the description itself (rather than those of the resource being described). Technical : groups the technical features of the resource. Educational : groups the educational and pedagogic features of the resource. Rights Management: groups features that depend on the kind of use envisaged for the resource. Table 1 resumes all categories and data elements that constitute ABITS Metadata Scheme. This is an IEEE LTSC LOM formally derived scheme. General MetaMetaData Educational RightsManagmnt Identifier Title Language Description Domain Idea Structure LifeCycle Version Create Create MetadataScheme Technical Format Size LocSpec Requirements Type Name MinimumVersion MaximumVersion PedagogicalType CoursewareGenre Format Approach InteractivityLevel SemanticDensity EducationalUse Role Difficulty Level Duration Role Description Conditions Reciprocity Attribution Prize MonetaryUnit Amount UnitOfPricing Table 1. ABITS Learning Object Metadata Scheme 2.2 Conceptual Graphs Metadata schemas not only have to provide information about a single Learning Object but they must also provide information about object relations and interdependency. For this purpose LOM standard provides an important data element called Idea (under the General category) that allows the so-called Domain Conceptualization. A Conceptualization is an abstract, simplified view of the world that we wish to represent for some purpose. A Conceptual Graph is an explicit specification of a Conceptualization. They can be viewed as structures of Concepts and conceptual relations where every arc links some conceptual relation r to some concept c [3]. With the term Concept we intend an abstract notion that refers to a particular Conceptual Graph. Within the ABITS context, Conceptual Graphs are used to link concepts underlying the knowledge domain with three kinds of relations: prerequisite, sub-concept and general relation (see table 2 for more details). Kind of Relation Relation Name Abbreviation Prerequisite Sub-Concept IsRequiredBy Requires IsPartOf HasPart IRB R IPO HP General Relation IsRelatedTo IRT Table 2. ABITS Concept Relations As an example of Conceptualization consider the concepts of addition, subtraction, multiplication, division and basic operations inside the domain of mathematical operations. Naturally in the Conceptual Graph of the Domain of Arithmetic, concepts must be represented as in figure 2. Division Basic Operations IRB HP Multiplication IRB Addition IRB Subtraction Fig. 2. An example of Conceptual Graph A standard to allow communication between systems that require a structured representation for logic is going to be defined. It is the Conceptual Graph Interchange Form (CGIF) [4] and has been developed as a conceptual schema language, as specified by ISO/IEC on Conceptual Schema Modeling Facilities (CSMF). We chose to represent ABITS Conceptual Graphs using this format. Such kind of information about the domain is massively used by ABITS functions in conjunction with Metadata fields for Cognitive State modeling and automatic Curriculum generation. While Metadata fields give information about Learning Object including explained Concepts (the already quoted Idea field), Conceptual Graphs give information about how Concept explained in these Learning Objects are related between themselves. In the following chapters we will see how such information is exploited by ABITS. 3 Student Modeling In every instant ABITS must be able to determine both Cognitive State and Learning Preferences for each student basing on his developed activities. Such set of information constitutes the so-called Student Model. Such structures, as we will see in the following two paragraphs, are composed by many fuzzy fields so it is useful now to provide a small Fuzzy Numbers overview. A Fuzzy Number is a concept related to the fuzzy set theory, an extension of the conventional set theory born in 1965 by the work of Zadeh [6]. The fuzzy set theory is dealt with subsets of an universe where the transition between the full membership and the non-membership is gradual rather than sudden. If X is an objects space, a fuzzy subset A of X is a set characterized by a membership function of the type: µ( x A): X 0, 1. (1) where x is a generic element of X and µ( x A ) is told membership degree of x in A. A subset in its classical meaning (crisp) can be then seen as a particular case of fuzzy subset with membership function with values in { 0,1 } where 1 indicate the full membership and 0 the non membership. A Fuzzy Number is a fuzzy subset of the set of real numbers with membership function µ( x A ) continuous, normal and convex able to satisfy the following requirements: x R such that µ( x A ) =1 (normality); (2) µ ( x A) min{ µ ( x1 A), µ ( x2 A)} x x1, x2 (convexity). (3) Exploiting fuzzy numbers it is possible to mathematically represent quantity of the kind: approximately 5 or few less than 3 and so on, in way to model uncertainty in intuitive manner and without resorting to the artifice of probability distributions. A fuzzy number can be graphically represented through the so-called belief graphs that map membership function for all support values (it is said support of a fuzzy set the set of all elements of the universe that have membership degree greater then 0). Figure 3 shows a belief graphs for a triangular fuzzy number where only a value has maximum membership degree Fig. 3. A Triangular Fuzzy Number 3.1 Cognitive State For Cognitive State it is intended the knowledge degree, reached by a particular student, of every Domain Concept (for each provided Domain). This information is logically represented by a string of fuzzy numbers (one for each concept) [8]. Graphically such string can be view as in figure 4. This string is updated dynamically by ABITS during training and testing activities. Fig. 4. A Graphical Representation of a Cognitive State The decision to use fuzzy numbers in Cognitive States arises from the necessity to manage uncertainty in the student evaluation. In this way, in fact, we can admit different kind of evaluations with different degree of reliability. For example, when a student reads an expositive Learning Object with a particular set of Concepts involved, ABITS can infer that there is a little increase in the student Cognitive State for these concepts but with a very large degree of uncertainty. When a student answer to a test in a correct way there is also an increase in his cognitive state in relation to involved concepts but with a more and more low degree of uncertainty. To represent this kind of information we use more and more narrow fuzzy numbers. An other important aspect to take into consideration is that a student could forget what he has learned some times before. In order to model the attitude that have humans to forget what they learn, ABITS applies a Forgetting Function to Cognitive States. Cause that not all humans forget in the same manner, this algorithm can t decrease knowledge degrees but only can wide the amplitude of representing fuzzy number signifying that evaluations are more and more unreliable. 3.2 Learning Preferences Within Learning Preferences we enclose all information about the student perceptive capabilities i.e. to which kind of resources a specified student is shown to be more receptive. To evaluate student preferences we exploit data elements contained into the Educational IEEE metadata category such as: format (kind of media), pedagogical approach, interactivity level, semantic density and difficulty [8]. Table 3 resumes all Learning Preferences fields with all possible values. Fields Format Approach Interactivity Level Semantic Density Difficulty Possible Values Text, Image, Slide, Hypertext, Video Clip, Simulation, Virtual Reality Inductive, Deductive, Explorative Very Low, Low, Medium, High, Very High Very Low, Low, Medium, High, Very High Very Low, Low, Medium, High, Very High Table 3. ABITS Learning Preferences Possible Values For the Format data element, ABITS maintains a list of fuzzy number where each one of them evaluates the receptiveness of a particular student to involved media (text, hypertext, video clip, simulation, virtual world, slide, etc.). The same idea is adopted in relation to pedagogical Approach field to evaluate if student is more interested in inductive, deductive or exploratory approach to learning. In the case of Interactivity Level, Semantic Density and Difficulty, represented by a single level value (very low, low, medium, high, very high), ABITS maintains a fuzzy number (for each parameter) representing the best receptive level for a particular student. In order to evaluate student Preferences, ABITS exploits this idea: during the learning process there are some Milestones (established by tutors) when the Cognitive State is updated with respect to activities performed by students. This means that for each Domain Concept involved in the student performed activities, a new evaluation is given. Exploiting the conceptual variation for each Concept and Metadata information on Learning Objects visited concerning that Concept, ABITS can evaluate the pedagogical effectiveness of Learning Object typologies. For example, if the knowledge of a particular Concept is sensibly raised between two milestones and visited Learning Objects concerning this particular Concept have been for a great part simulations, ABITS infers that the student is receptive to simulations. The system increases consequently the format Preference that refers to simulations. The information calculated by ABITS about Student Models can be exploited directly by tutors or can be re-used by ABITS in the automatic Curriculum generation procedure. In the following chapter we will explain how ABITS can do that. 4 Automatic Curriculum Generation Each student can be assigned to one or more different Courses. An ABITS Course is composed essentially by an set of Learning Goals and by a Curriculum [8]. With Learning Goals (that are strongly different from Learning Objects) we intend a set of key Concepts necessary to be learned to successful complete a specific Course. Such Concepts (as all other Concepts) are part of a Domain and are represented inside the Conceptual Graph of such Domain. With Curriculum we intend, instead, an ordered list of Learning Objects that can be used to provide to a specific student all necessary knowledge to complete a specific Course. While Learning Goals indicate what (which Concepts) a student has to learn, Curriculum specify how these Concepts has to be learnt. Different students can require different Curriculums to learn about the same Learning Goals depending on their Cognitive States and Learning Preferences. For this reason a Curriculum Generation procedure is also provided by ABITS. Starting from the list of Learning Goals that must be covered, ABITS Curriculum Generation is done by a three-step procedure [9]. 1. The list LG of Learning Goals is exploded in order to obtain the list C of all Concepts that a student must learn to reach such Goals. This is obtained using a recursive function: for each concept c contained in LG a test is done: if c is already known by the student (looking his Cognitive State) then c is discarded, otherwise the c is added to C and all c requisites are kept (looking the Conceptual Graph of the c Domain) and added to LG. The procedure is repeated until LG is empty. 1. The list of concept C is transformed in a list of Learning Objects LO finding the best ordered sequence of Learning Object from the Course Material Database matching student Learning Preferences (comparing Metadata Educational Fields with student Preferences fields) and covering C. LO must be ordered: if concept c1 requires concept c 2 then any Learning Object explaining c 1 will be placed in the Curriculum after any Learning Object explaining c The list of Learning Objects LO is transformed in a Curriculum CURR by adding Testing Material and Milestones. Testing material are simply Testing Learning Objects (i.e. HTML pages with evaluation forms) while Milestones are embedded ABITS calls. Milestones are placed after each testing block and at the end of the whole curriculum in order to advice the Course Management System that ABITS must be called in this place to update the model and/or the Curriculum sequence for a specified student. It is important to note that all ABITS functions could be used by the Course Management System in a systematic or occasional manner according to tutors and administrators policies. Information generated in the student modeling phase in-fact can be used directly by ABITS in the Curriculum generation process or
Search
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks